HARU-Net: Hybrid Attention Residual U-Net for Edge-Preserving Denoising in Cone-Beam Computed Tomography
#HARU-Net #Cone-Beam Computed Tomography #Edge-Preserving Denoising #Deep Learning #Medical Image Processing #Low-Dose Imaging #Hybrid Attention Transformer #Diagnostic Quality
📌 Key Takeaways
- HARU-Net effectively denoises CBCT images while preserving critical anatomical edges
- The method outperforms existing techniques with superior PSNR, SSIM, and GMSD metrics
- HARU-Net achieves high performance at lower computational cost than current methods
- The approach addresses the scarcity of high-resolution CBCT data for training
- The research focuses on improving diagnostic quality in low-dose CBCT imaging
📖 Full Retelling
Researchers Khuram Naveed and Ruben Pauwels introduced HARU-Net, a Hybrid Attention Residual U-Net for edge-preserving denoising in Cone-Beam Computed Tomography, in a paper submitted to arXiv on February 26, 2026, addressing the critical challenge of maintaining diagnostic image quality in low-dose CBCT scans commonly used in dental and maxillofacial imaging. The research tackles a significant problem in medical imaging where low-dose acquisition introduces strong, spatially varying noise that degrades soft-tissue visibility and obscures fine anatomical structures. Classical denoising methods struggle to suppress noise while preserving edges, and existing deep learning approaches face limitations due to the scarcity of high-resolution CBCT data for supervised training. To overcome these challenges, the researchers developed HARU-Net, trained on a cadaver dataset of human hemimandibles using a high-resolution protocol from the 3D Accuitomo 170 CBCT system. The innovative approach integrates three complementary architectural components: hybrid attention transformer blocks within skip connections to emphasize salient anatomical features, residual hybrid attention transformer groups at the bottleneck to strengthen global contextual modeling, and residual learning convolutional blocks for deeper, more stable feature extraction. Performance metrics demonstrate that HARU-Net consistently outperforms state-of-the-art methods including SwinIR and Uformer, achieving the highest PSNR (37.52 dB), highest SSIM (0.9557), and lowest GMSD (0.1084) while maintaining significantly lower computational costs, offering a practical solution for improving diagnostic quality in low-dose CBCT imaging.
🏷️ Themes
Medical Imaging, Artificial Intelligence, Image Processing
📚 Related People & Topics
Deep learning
Branch of machine learning
In machine learning, deep learning focuses on utilizing multilayered neural networks to perform tasks such as classification, regression, and representation learning. The field takes inspiration from biological neuroscience and revolves around stacking artificial neurons into layers and "training" t...
Entity Intersection Graph
Connections for Deep learning:
View full profileOriginal Source
--> Electrical Engineering and Systems Science > Image and Video Processing arXiv:2602.22544 [Submitted on 26 Feb 2026] Title: HARU-Net: Hybrid Attention Residual U-Net for Edge-Preserving Denoising in Cone-Beam Computed Tomography Authors: Khuram Naveed , Ruben Pauwels View a PDF of the paper titled HARU-Net: Hybrid Attention Residual U-Net for Edge-Preserving Denoising in Cone-Beam Computed Tomography, by Khuram Naveed and Ruben Pauwels View PDF HTML Abstract: Cone-beam computed tomography is widely used in dental and maxillofacial imaging, but low-dose acquisition introduces strong, spatially varying noise that degrades soft-tissue visibility and obscures fine anatomical structures. Classical denoising methods struggle to suppress noise in CBCT while preserving edges. Although deep learning-based approaches offer high-fidelity restoration, their use in CBCT denoising is limited by the scarcity of high-resolution CBCT data for supervised training. To address this research gap, we propose a novel Hybrid Attention Residual U-Net (HARU-Net) for high-quality denoising of CBCT data, trained on a cadaver dataset of human hemimandibles acquired using a high-resolution protocol of the 3D Accuitomo 170 (J. Morita, Kyoto, Japan) CBCT system. The novel contribution of this approach is the integration of three complementary architectural components: a hybrid attention transformer block embedded within each skip connection to selectively emphasize salient anatomical features, a residual hybrid attention transformer group at the bottleneck to strengthen global contextual modeling and long-range feature interactions, and residual learning convolutional blocks to facilitate deeper, more stable feature extraction throughout the network. HARU-Net consistently outperforms state-of-the-art methods including SwinIR and Uformer, achieving the highest PSNR (37.52 dB), highest SSIM (0.9557), and lowest GMSD (0.1084). This effective and clinically reliable CBCT denoising is achieved at a ...
Read full article at source